Support masks in audio metrics

Nobody has claimed this yet.

## 🚀 Feature

It would be great if [ScaleInvariantSignalDistortionRatio](https://torchmetrics.readthedocs.io/en/stable/audio/scale_invariant_signal_distortion_ratio.html#scale-invariant-signal-to-distortion-ratio-si-sdr) and [ScaleInvariantSignalNoiseRatio](https://torchmetrics.readthedocs.io/en/stable/audio/scale_invariant_signal_noise_ratio.html) allowed an extra argument `mask` when computing a metric on a batch of audio frames which have been zero-padded to fit the batch. Allowing to pass a mask would only compute the metric on valid audio frames across the batch.

### Alternatives

In the TorchAudio [tutorial](https://github.com/pytorch/audio/blob/main/examples/source_separation/utils/metrics.py) they have manually computed the metrics to account for this.
Would be call if torchmetrics did this.

Start with the implementations of ScaleInvariantSignalDistortionRatio and ScaleInvariantSignalNoiseRatio, then compare the manual masked calculations in the linked TorchAudio tutorial. Done means both metrics accept a mask and exclude zero-padded audio frames when computing batch results, with behavior covered by appropriate tests.

Written by the indexing model from the issue text.